<-- Go back

Improved sampling via learned diffusions

by Lorenz Richter, Julius Berner, Guan-Horng Liu




eprint arXiv:2307.01198


Recently, a series of papers proposed deep learning-based approaches to sample from unnormalized target densities using controlled diffusion processes. In this work, we identify these approaches as special cases of the Schrödinger bridge problem, seeking the most likely stochastic evolution between a given prior distribution and the specified target. We further generalize this framework by introducing a variational formulation based on divergences between path space measures of time-reversed diffusion processes. This abstract perspective leads to practical losses that can be optimized by gradient-based algorithms and includes previous objectives as special cases. At the same time, it allows us to consider divergences other than the reverse Kullback-Leibler divergence that is known to suffer from mode collapse. In particular, we propose the so-called log-variance loss, which exhibits favorable numerical properties and leads to significantly improved performance across all considered approaches.


Read the paper

Additional Information

Brief introduction of the dida co-author(s) and relevance for dida's ML developments.

About the Co-Author

With an original focus on stochastics and numerics (FU Berlin), the mathematician has been dealing with deep learning algorithms for some time now. Besides his interest in the theory, he has practically solved multiple data science problems in the last 10 years. Lorenz leads the machine learning team.